**Background:** Inflammatory bowel disease (IBD) in children is associated with malnutrition, growth restriction, and altered body composition. Monitoring nutritional status is critical for disease management, but dual-energy X-ray absorptiometry (DEXA), the clinical gold standard for body composition assessment, is not always available. Existing predictive equations for fat mass percentage (FM%) were developed for the general population and may not be accurate in pediatric IBD. This study aimed to compare body composition between IBD patients and healthy controls (HCs), evaluate the accuracy of anthropometric measures for predicting FM%, and develop population-specific formulas for FM% estimation in children with IBD.
**Methods:** Thirty IBD patients (median age 14 years, IQR 11–16; 16 CD, 12 UC, 2 IBD-U) were prospectively recruited from the Gastroenterology Unit of Vittore Buzzi Children's Hospital, Milan, between September 2019 and May 2020. Inclusion criteria were age 6–18 years and confirmed IBD diagnosis. Exclusion criteria were age <6 or >18 years and diagnosis under definition. A control group of 144 healthy children and adolescents (median age 15 years, IQR 14–17) was recruited from the International Center for the Assessment of Nutritional Status (ICANS, University of Milan). All participants underwent anthropometric measurements (weight, height, BMI, skinfold thicknesses at biceps, triceps, subscapular, and suprailiac sites) and DEXA scanning (Lunar Prodigy Advance, GE Medical Systems). FM% was calculated from DEXA. Four prespecified linear regression models were used to predict FM% from: Model 1 (logeBMI), Model 2 (loge triceps skinfold [TSF]), Model 3 (loge sum of 2 skinfolds [biceps+triceps, SF2]), and Model 4 (loge sum of 4 skinfolds [biceps+triceps+subscapular+suprailiac, SF4]). All models were adjusted for sex, age, IBD status, and the interaction between the anthropometric predictor and IBD status. Model fit was assessed using adjusted R² and root mean squared error (RMSE) with 95% confidence intervals calculated via bootstrap on 1000 samples. Bland-Altman analysis was used to compare existing predictive equations (Brook, Johnston, Durnin & Rahaman, Durnin & Womersley) against DEXA-measured FM%.
**Key Results:** IBD patients had significantly lower weight (median 51 kg vs 60.5 kg, p=0.004), BMI (median 20.1 kg/m² vs 22.4 kg/m², p=0.002), and absolute fat mass (median 12.4 kg vs 18.3 kg, p=0.002) compared to HCs. However, FM% did not differ significantly between groups (29.6% vs 32.2%, p=0.108). All skinfold thicknesses were significantly lower in IBD patients except biceps skinfold. The regression models showed: Model 1 (BMI) had R²=0.68 and RMSE=4.94%; Model 2 (TSF) had R²=0.82 and RMSE=3.75%; Model 3 (SF2) had R²=0.86 and RMSE=3.23%; Model 4 (SF4) had R²=0.88 and RMSE=3.03%. The interaction between anthropometric predictor and IBD status was not significant in any model, indicating parallel regression lines for IBD and control groups. Among existing equations, the Brook equation had the lowest median percentage bias (−5.6%, IQR −17.5% to 3.3%), while Durnin & Womersley had the highest (−26.4%, IQR −36.7% to −16.4%). Bland-Altman plots revealed proportional bias affecting all existing equations (Pitman test p<0.05). The newly developed population-specific formulas are provided for clinical use.
**Clinical Implications:** This study provides validated, population-specific anthropometric formulas for estimating FM% in children with IBD, offering a practical alternative when DEXA is unavailable. The sum of 4 skinfolds (triceps, biceps, subscapular, suprailiac) is the most accurate predictor (R²=0.88, RMSE=3.03%), while the sum of 2 skinfolds (triceps+biceps) offers slightly lower accuracy (R²=0.86, RMSE=3.23%) but greater feasibility and less susceptibility to measurement error. BMI alone is a poor predictor of body composition (R²=0.68). These formulas can facilitate routine nutritional monitoring in pediatric IBD clinics. Limitations include the small sample size (n=30 IBD patients), which precluded subgroup analysis by disease type (CD vs UC), and the lack of adjustment for disease activity and treatment effects. Further prospective studies are needed to validate these formulas in larger, independent cohorts.